Table 2.

Model Hyperparameters

	TTB	TIPS	UAE
SMOTE oversampling percentage	300%	1,000%	800%
Total number of features in random forest	95	81	58
Total number of trees in random forest	1,000	1,000	1,000
Number of features per individual tree	10	9	8

Note–Hyperparameters for each of the 3 random forest models. Hyperparameters describe the process used for constructing the random forest models. The SMOTE oversampling percentage describes how many synthetic data points were created from the original dataset. For example, a SMOTE oversampling percentage of 300% (respectively, 1,000% and 800%) means that for each patient in the original dataset who had the outcome of interest, 3 (respectively, 10 and 8) synthetic data points were created. Each random forest consisted of 1,000 trees. Each individual tree was trained on a subset of features (selected randomly) and a subset of patients (70% of the total patients, selected randomly). There was no maximum node depth (ie, there was no limit to the size of any individual tree).

SMOTE = Synthetic Minority Oversampling Technique; TIPS = transjugular intrahepatic portosystemic shunt; TTB transthoracic biopsy; UAE = uterine artery embolization.